Kurtis D. Cantley

Graphene Foam as a Three-Dimensional Platform for Myotube Growth

This study demonstrates the growth and differentiation of C2C12 myoblasts into functional myotubes on 3-dimensional graphene foam bioscaffolds. Specifically, we establish both bare and laminin coated graphene foam as a biocompatible platform for muscle cells and identify that electrical coupling stimulates cell activity. Cell differentiation and functionality is determined by the expression of myotube heavy chain protein and Ca2+ fluorescence, respectively. Further, our data show that the application of a pulsed electrical stimulus to the graphene foam initiates myotube contraction and subsequent localized substrate movement of over 100 micrometers. These findings will further the development of advanced 3-dimensional graphene platforms for therapeutic applications and tissue engineering.

Signal-to-noise ratio enhancement using graphene-based passive microelectrode arrays

This work is aimed toward the goal of investigating the influence of different materials on the signal-to-noise ratio (SNR) of passive neural microelectrode arrays (MEAs). Noise reduction is one factor that can substantially improve neural interface performance. The MEAs are fabricated using gold, indium tin oxide (ITO), and chemical vapor deposited (CVD) graphene. 3D-printed Nylon reservoirs are then adhered to the glass substrates with identical MEA patterns. Reservoirs are filled equally with a fluid that is commonly used for neuronal cell culture. Signals are applied to glass micropipettes immersed in the solution, and response is measured on an oscilloscope from a microprobe placed on the contact pad external to the reservoir. The time domain response signal is transformed into a frequency spectrum, and SNR is calculated from the ratio of power spectral density of the signal to the power spectral density of baseline noise at the frequency of the applied signal. We observed as the magnitude or the frequency of the input voltage signal gets larger, graphene-based MEAs increase the signal-to-noise ratio significantly compared to MEAs made of ITO and gold. This result indicates that graphene provides a better interface with the electrolyte solution and could lead to better performance in neural hybrid systems for in vitro investigations of neural processes.

Modeling Memristor Radiation Interaction Events and the Effect on Neuromorphic Learning Circuits

An ideal memristor model is modified to include the effects of radiation interactions with the device. Modeling is done in Cadence Virtuoso design suite using Verilog-A. Simulations include the effect of radiation events that could change the state of device or can ionize the device to create e-h+ pairs or change the off-state resistance of the device. Combination of these events occurring simultaneously is also studied. Simulation results are compared with the experimental results published in existing research papers. Finally, transient simulation of a three-input, two-output spiking electronic neural network with memristive synapses is performed. Varying amounts of energy deposited by radiation are modeled, and it is observed that radiation exposure dramatically alters the synaptic weight evolution.

Electrical characteristics of nanocrystalline silicon resistive memory devices

Resistive memory devices have been studied and fabricated using a wide variety of materials including chalcogenides, metal oxides, and hydrogenated amorphous silicon (a-Si:H). The most promising materials seem to be amorphous in nature, with the properties of the atomic lattices being conducive to the physical mechanisms that underlie the subsequent resistive switching. The devices are also finding applications beyond high-density digital memory, such as for electronic synapses in neuromorphic systems. However, a different set of properties is required in the latter case compared to devices that must only store binary values. In addition, it is well known that biological synapses are extremely unreliable and noisy, and yet the brain is still able to perform high-level cognitive functions. This work uses pulse-based electrical characterization techniques to demonstrate the stochastic nature of resistive switching in nanocrystalline silicon (nc-Si) Conductive Bridge Resistive Memory (CBRAM) Devices. We chose nc-Si active layers so these devices could potentially be co-fabricated in the same process as nc-Si TFTs. Our subsequent findings indicate the device properties may be particularly useful for some non-von Neumann computing paradigms. Though much research has been done using a-Si:H, results from nc-Si CBRAM devices have not been published. In this study, we showed that the switching of the device depends on the history of current passing though it, and not only the voltage applied. Further, the resistance switching in the devices is stochastic, making them ideal candidates for a biologically realistic synapse.

Spatio-temporal pattern recognition in neural circuits with memory-transistor-driven memristive synapses

Spiking neural circuits have been designed in which the memristive synapses exhibit spike timing-dependent plasticity (STDP). STDP is a learning mechanism where synaptic weight (the strength of the connection between two neurons) depends on the timing of pre-and post-synaptic action potentials. A known capability of networks with STDP is detection of simultaneously recurring patterns within the population of afferent neurons. This work uses SPICE (simulation program with integrated circuit emphasis) to demonstrate the spatio-temporal pattern recognition (STPR) effect in networks with 25 afferent neurons. The neuron circuits are the leaky integrate-and-fire (I&F) type and implemented using extensively validated ambipolar nano-crystalline silicon (nc-Si) thin-film transistors (TFT) models. Ideal memristor synapses are driven by a nanoparticle memory thin-film transistor (np-TFT) with a short retention time attached to each neuron circuit output. This device serves to temporally modulate the conductance path from post-synaptic neurons, providing rate-based and timing-dependent learning. With this configuration, the use of a crossbar structures would also be possible, providing dense synaptic connections and potentially reduced energy consumption.

A CMOS synapse design implementing tunable asymmetric spike timing-dependent plasticity

A CMOS synapse design is presented which can perform tunable asymmetric spike timing-dependent learning in asynchronous spiking neural networks. The overall design consists of three primary subcircuit blocks, and the operation of each is described. Pair-based Spike Timing-Dependent Plasticity (STDP) of the entire synapse is then demonstrated through simulation using the Cadence Virtuoso platform. Tuning of the STDP curve learning window and rate of synaptic weight change is possible using various control parameters. With appropriate settings, it is shown the resulting learning rule closely matches that observed in biological systems.